ext4_insert_range() grows the file by len bytes but only checks the new size against s_maxbytes. FALLOC_FL_INSERT_RANGE can therefore make a file larger than RLIMIT_FSIZE allows, while ftruncate() and the fallocate() modes that extend a file fail with -EFBIG and SIGXFSZ. xfs, f2fs and ntfs3 already refuse to insert a range past the limit. Fix this by validating the new size with inode_newsize_ok(), as ext4_do_fallocate() and ext4_zero_range() do. Fixes: 331573febb6a ("ext4: Add support FALLOC_FL_INSERT_RANGE for fallocate") Cc: stable@vger.kernel.org Signed-off-by: Hungyu Lin --- fs/ext4/extents.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c index ad0d80e33f60..94b9e091ef5f 100644 --- a/fs/ext4/extents.c +++ b/fs/ext4/extents.c @@ -5803,6 +5803,9 @@ static int ext4_insert_range(struct file *file, loff_t offset, loff_t len) /* Check whether the maximum file size would be exceeded */ if (len > inode->i_sb->s_maxbytes - inode->i_size) return -EFBIG; + ret = inode_newsize_ok(inode, inode->i_size + len); + if (ret) + return ret; /* * Persist the pending zeroed EOF block to ensure i_disksize -- 2.43.0